Collection and linguistic processing of a large-scale corpus of medical articles

We have collected a large-scale corpus of electronic articles in the cardiology domain (85 million+ words) in the framework of a digital library project that tailors the presentation of online medical literature to both patients and healthcare providers. We describe the webbased and XML technologies we used for the collection, encoding and linguistic processing of the corpus. This resulted in a largescale, high-quality, thoroughly marked-up resource which is used by many researchers in our project, in the areas of natural language processing, information retrieval and medical informatics. We show how the final use of the resource has influenced the design of its structural and linguistic encoding. The procedure we describe is general enough to be of use to researchers in a similar position wishing to compile, encode and linguistically annotate their own corpus from the web.